Machine learning to help identify “hidden fat” on routine bone scans

Researchers from Edith Cowan University (ECU) are developing an artificial intelligence algorithm that can use bone density scans captured to detect potential spine fractures to estimate visceral fat levels, providing a quick, painless and affordable solution.

“Ever heard of the sneaky fat that hides deep inside your belly and wraps around your organs? That’s visceral fat - a real troublemaker that is strongly linked to serious health problems like heart disease, diabetes, and cancer,” said PhD student Ms Arooba Maqsood.

“Obesity poses a serious threat to global health and is a leading cause of morbidity and mortality worldwide. Beyond its toll on health, the economic burden is staggering, placing immense strain on healthcare systems and national economies alike.

“For Australia, the economic cost was A$39 billion in 2019, projected to reach A$228 billion by 2060 which is 3.5 per cent of Australia’s gross domestic product. And it’s not just about money. It is estimated there are 3.7 million obesity-related deaths each year globally,” Ms Maqsood said.

The dangerous visceral fat wrapped around your organs is currently estimated using methods such as body mass index, waist circumference, and waist-to-hip ratio. However, Ms Maqsood noted that these measures have limitations, and do not distinguish between different types of body fats.

“This oversimplification contributes to inconsistencies in assessment of obesity and its complications, highlighting the need for a more precise approach to measure obesity,” she added.

“Although imaging techniques like MRI and CT scans can accurately measure visceral fat, their hefty price tag is often a limiting factor. Whilst CT also exposes patients to higher levels of radiation.”

Lateral spine Dual-energy X-ray Absorptiometry (DXA) scans are used to identify spine fractures. These images can be repurposed for opportunistic screening for visceral fat, giving us new and valuable health insights without the need for extra tests.

ECU is training its machine learning algorithm to use on these scans to accurately predict the amount of visceral fat present in a person simply by looking at lateral spine DXA scans.

“The machine-learning model has been trained on thousands of images; the next step is to incorporate further datasets from around the world, so it learns from the largest, most diverse cohort possible and becomes as effective as possible,” said Dr. Syed Zulqarnain Gilani, a senior lecturer and lead AI scientist at ECU.

“The main work of our research group is focused on early intervention and is supported by the Heart Foundation, Raine Medical Research Foundation and WA Department of Health through the Future Health Research and Innovation (FHRI) Fund,” Dr Syed Zulqarnain Gilani added.

Ms Maqsood will be presenting her research at the International Conference on Medical Image Computing and Computer Assisted Interventions in Korea in September this year.